ocr-space
Extract text from images or PDF documents using the OCR.space API. Use this skill whenever the user needs to recognize text from an image URL, a local image file, or a Base64-encoded image, including
SKILL.mdresources/migrated/v1-skills/ocr-space/SKILL.md
GitHubOCR.space — Image and PDF Text Recognition
Capabilities Overview
Perform optical character recognition (OCR) on images and PDF files via the OCR.space API. Supports 30+ languages, three submission methods (URL, file upload, Base64), and can return recognized text along with coordinate information.
This plugin contains two endpoints:
| Endpoint | Method | Description | Billing |
|---|---|---|---|
POST https://app-dohnbj5by5mp-api-W9z3M6eONl3L.gateway.appmedo.com/parse/image |
POST | Core OCR endpoint supporting URL / file upload / Base64 input methods, with advanced features such as table recognition and coordinate overlay | Billed per call |
GET https://app-dohnbj5by5mp-api-m9xKXDbRplNa.gateway.appmedo.com/parse/imageurl |
GET | Simplified GET endpoint supporting URL-only image submission, suitable for quick integration | Not billed |
Prefer the GET /parse/imageurl endpoint (free) for simple URL-based image recognition. Use the POST /parse/image endpoint only when you need file upload, Base64 input, table recognition, coordinate overlay, or other advanced features.
End-to-End Workflow
Both endpoints are synchronous — no polling required: submit the request and receive the recognition result directly.
User provides image source
↓
Determine input type
├─ Image URL (no advanced needs) → GET /parse/imageurl (free, simple, fast)
└─ File upload / Base64 / advanced features → POST /parse/image (per-call billing)
↓
Parse ParsedResults[0].ParsedText from the response
↓
Return recognized text (or coordinate information)
Generation-Time Usage (Agent Direct Call)
Use the built-in script for generation-time calls. The script reads INTEGRATIONS_API_KEY from the environment.
The Bash tool timeout MUST be set to 600000ms (600 seconds).
# OCR from an image URL (preferred — not billed)
python3 <skill-path>/scripts/call_ocr_space.py --url "https://example.com/document.jpg" --language eng
# OCR from a local file (billed)
python3 <skill-path>/scripts/call_ocr_space.py --file /path/to/document.png --language eng --engine 2
# With table recognition and scaling
python3 <skill-path>/scripts/call_ocr_space.py --url "https://example.com/table.jpg" --table --scale
The script prints {"status":"succeed","result":{...}} on success. On failure it prints an error to stderr and exits with a non-zero code.
See references/parse-image-api.md (POST, supports file/Base64) and references/parse-imageurl-api.md (GET, URL only) for the full parameter tables.
Post-Generation Usage (Application Calls via Edge Function)
See the Edge Function boilerplate code in each endpoint reference file:
- POST endpoint Edge Function → read
references/parse-image-api.md - GET endpoint Edge Function → read
references/parse-imageurl-api.md
Security contract:
- The frontend sends the image source (URL / Base64) to the Edge Function (JSON body).
- The Edge Function reads the API key from the Deno environment variable
INTEGRATIONS_API_KEYand adds it to theX-Gateway-Authorizationheader. - The raw API Key is never exposed to the browser.
- 429 (quota exceeded) and 402 (insufficient balance) errors are forwarded verbatim to the frontend.
Notes
- Key Security:
INTEGRATIONS_API_KEYcan only be read server-side in the Edge Function; never expose it to the frontend. - Error Handling: Always handle 429 (quota exceeded) and 402 (insufficient balance); also check
OCRExitCode(1=success, 2=partial success, 3=failure, 4=error) andIsErroredOnProcessingin the response body. - Billing: POST
/parse/imageis billed per call, but unnecessary repeated calls should still be avoided. GET/parse/imageurlis not billed and should be preferred. - Input Constraints: The POST endpoint requires one of
url,file, orbase64Image; Base64 input must include the data type prefix, e.g.data:image/jpeg;base64,. - Language Codes: Use OCR.space proprietary language codes (e.g.
chsfor Simplified Chinese,chtfor Traditional Chinese), not standard BCP 47 codes.